Taking a balanced approach to remote patient monitoring
The medical Internet of Things is filled with many valuable devices--and a few questionable ones. It’s critically important to separate clinically validated tools from marketing hype.
By John Halamka, M.D., president, Mayo Clinic Platform, and Paul Cerrato, senior research analyst and communications specialist, Mayo Clinic Platform
When William Shakespeare penned the words “All that glitters is not gold,” it’s unlikely he had the Internet of Things (IoT) in mind. And yet that truism perfectly describes the situation many clinicians and patients find themselves in as they attempt to choose reliable remote monitoring tools. The potential pitfall is well illustrated by a recent warning letter sent by the US Food and Drug Administration (FDA) to an IoT device vendor making unsubstantiated claims. On the other hand, a review of several controlled clinical trials demonstrates that some devices are in fact “golden.” Consider a few examples.
A company that makes ‘smart socks’ intended to be put on a newborn’s foot to monitor oxygen level, heart rate, and total hours slept received a warning from the FDA saying the company has been making unapproved claims. According to the warning letter, the company’s sales team stated: “We look at the best indicators of your baby’s overall well-being and will proactively notify you if your baby may need you.” FDA explains that products like this are “devices because they are intended for use in the diagnosis of disease or other conditions or in the cure, mitigation, treatment, or prevention of disease, or to affect the structure or any function of the body. Products that measure blood oxygen saturation and pulse rate are devices when they are intended to identify (diagnose) desaturation and bradycardia and provide an alarm to notify users that measurements are outside preset values.” And since the vendor never submitted the necessary documentation to the agency to demonstrate its product is safe and effective medical device, it’s in violation of federal law.
At the other end of the IoT spectrum are devices like IDx-DR, AliveCor, and several other well-researched products, all of which have FDA clearance or approval. Researchers from Mayo Clinic and AliveCor Inc. have been using artificial intelligence (AI) to develop a mobile device that can identify certain patients at risk of several heart conditions, including atrial fibrillation that can lead to stroke. The researchers determined that a smartphone-enabled mobile EKG device can rapidly and accurately determine a patient's QTc, thereby identifying patients at risk. In addition, FDA allowed the use of KardiaMobile 6L to measure QTc in COVID-19 Patients to monitor QT duration in patients receiving medications that can cause potentially life-threatening QT prolongation.
These innovations are grounded in in peer reviewed science. The European Journal of Internal Medicine has published a prospective, blinded evaluation of the AliveCor system for monitoring atrial fibrillation, comparing it to a standard 12-lead ECG and reported: “The AliveCor Kardia ECG monitor allows a highly accurate detection of atrial fibrillation by an interpreting electrophysiologist both in the standard lead I and a novel parasternal lead.” Similarly, a systematic review and meta-analysis of handheld EKG devices for atrial fibrillation screening concluded: “The pooled sensitivity and specificity of single-lead handheld ECG devices were high.” Prospective studies and meta-analyses are among the strongest forms of evidence to support a medical intervention.
The AliveCor KardiaMobile 6 lead device is designed to take a 6-lead reading, which requires patients to place two fingers on the top of the device and then place the bottom of the device on the bare skin of the left leg. The AliveCor single lead unit, on the other hand, only requires the 2 fingers be placed on the device. One approach is to have the patient place their thumb and forefinger on the unit for about 20 seconds, while the signal is recorded by the app running on a tablet. The data can then be sent to a hospital’s cloud server, where the QTc algorithm is executed against the signal data. QTc refers to heart rate–corrected QT interval; a prolonged QT interval on an EKG can help diagnose various types of heart disease, including atrial fibrillation that can lead to stroke or heart failure, if left undiagnosed.
At Mayo Clinic, the generated insight is then returned to a laptop running the Mayo CV developed QTc Tracker software for display and consultation. Once the signal is captured, it takes about 10 seconds to get the result back. The process is illustrated in the figure below.
No doubt there are many ethical IT developers who would like to see the success of AliveCor, IDx-DR and other well-researched products but find the barriers to clinical effectiveness and regulatory approval more than a little intimidating. Initially, developers need to collect an adequate number of potential users or patients to serve as a data set on which to train their algorithm; the data set needs to be representative of the real-world population that will eventually use it. That data set also needs to be unbiased, normalized, secure, and de-identified to remain compliant with HIPAA regulations. Depending on the intended function of the algorithm, the data set should include a combination of structured and unstructured data from a trustworthy EHR system. Once all the relevant data are collected, the trained algorithm needs to be cross-tested on independent data sets to ensure its generalizability. Finally, assuming the product has managed to jump the regulatory hurdles and rigor of the peer review process of medical thought leaders, it has to be seamlessly integrated into clinicians’ workflow. If they have to jump through too many hoops to use the algorithm, it’s unlikely it will attract enough users to justify the expense of installing the tool. With these concerns in mind, Mayo Clinic Platform has created four AI product development tools: Gather, Discover, Validate, and Deliver. Among the many advantages of using such platform services: They can address the lack of interoperability between health care systems, accept data and diagnostic signals from various sources—wearables, EHRs, clinical systems, diagnostic devices, etc.—as well as integrate, harmonize and store the data so it can seamlessly be used to advance innovative solutions.
The medical Internet of Things continues to evolve as telemedicine services become more widely accepted across the globe. But for patients and clinicians to accurately interpret the data generated by these tools, they need to be assured that these tools “glitter”—in the best sense of the word.